a group of people standing next to each other — Photo by Robynne O on Unsplash

Last week, over 700 of you showed up for the AI for Author Business Summit. I had planned to pre-record some sessions to save my voice, but Anthropic kept shipping updates faster than I could finish a slide deck, so we did the whole thing live. My voice survived. Barely. Theater kids don’t quit, y’all — the show goes on.

By the end of Day One, people were connecting Airtable to Claude Cowork and chatting about their 300+ backlist and series bible details. By Day Two, attendees were chatting with their email, setting up scheduled tasks to update their websites, and a few brave souls had hacked it to update their Scrivener docs, and clean their Mailerlite email lists. Someone else hooked up their Facebook Ads and rebuilt their entire strategy with new images and video they created. (I love you people.)

And then there were the social media shifts. They learned how to connect AuthorAutomations.social inside Claude Cowork and create and schedule all their posts from a chat window —including images, videos, infographics, carousels, and yes, Trial Reels (which seems to be moving the needle, according to some authors chatting about it in Wide for the Win!)

The shift we talked about all week was the move from chatting with AI to having AI take action on your behalf — with your permission, your guardrails, and your rules. That’s what agentic means, and it landed for a lot of people. Let me remind you if feel like you’re behind — only 2% of the world is actually using AI in their business. 2%. You’re officially an early adopter living in the future.

I’m going to break a lot of what we covered in this newsletter over this series, but if you want the cheat code (and save weeks of waiting for me to write) then replay packages for all 15 sessions + access to Discord are available at https://aisummit.indieauthortraining.com.

We also have three things coming up on the calendar. On April 29 at 11am Central, I’m doing an overview of StorytellerOS for anyone who wants to see the full platform in action. On May 5 from 1-3pm Central, we’re running a hands-on session on Claude Cowork for Ads Management. And every Wednesday at 12 noon Central, I’m hosting an Author Automations Social Getting Started session if you want help setting up the social scheduler and connecting it to your workflows.

Register for any or all of the webinars at https://webinars.indieauthortraining.com

Last week we broke down the differences between Claude Chat, Cowork, and Code.

One of the biggest questions after that was a version of “okay, I get that there are three Claudes, but where do I actually start?” This newsletter is the answer. We’re going to walk through the differences between the desktop app and the web version, talk about what a terminal is (for those of you who need it), and then set up the two concepts that make Claude Cowork and Claude Code genuinely powerful: Skills and Connectors. Next week, we’ll go deeper into both of those.

Desktop App vs. Web: Start With the Desktop

If you only take one thing from this newsletter, make it this: download the Claude desktop app and use that instead of the website, if you can.

The web version at claude.ai is fine for chatting. You can ask questions, paste in text, get answers — it works the same way it always has. But when you move to Cowork (the part where Claude can actually do things on your computer), the desktop app is where everything lives. Cowork in the desktop app can read files and folders on your machine, create and edit documents in place, connect to your tools through connectors, run scheduled tasks, build “live artifacts” (which are dashboards, but better), and access the skills you build to teach it how you work.

The web version can’t do most of that because it doesn’t have access to your local files. It runs in what’s called a sandbox up in the cloud, which means it can’t see your desktop, your documents folder, or any of the configuration you’ve done on your machine. Your skills don’t transfer to the web automatically either — you can manually set them up there, but it’s not reliable and you’ll spend more time troubleshooting than actually getting work done.

During the summit, someone asked about the pros and cons of desktop vs. web, and my answer was pretty simple: if you’re just chatting, either one works. If you want Claude to actually do things for you, desktop app every time.

To get it, go to claude.ai, look in the bottom left corner for “Get apps and extensions,” and download it for your Mac or PC. You’ll want at least the $20/month Pro plan to access Cowork, and if you end up being a power user (you’ll know when you get there), the $100/month plan gives you a lot more capacity.

About That Desktop App I Used to Hate

I need to make a confession here because some of you have been reading this newsletter long enough to remember that I was not exactly a fan of the Claude desktop app a month ago, and I didn’t recommend Cowork at all. It was buggy. Cowork was unimpressive. I had real concerns about whether Anthropic was going to get this right.

They got it right.

The improvement in the last four to six weeks has been dramatic, and I’m not using that word casually. The desktop app went from something I tolerated to something I rely on for hours a day. I see the “Relaunch to update” notification on my screen so often that I’m starting to wonder if Anthropic’s engineering team sleeps.

They are shipping updates at a pace I have never seen from a software company in nearly three decades of working in technology, and every update makes it noticeably better.

If you tried the desktop app a month ago and walked away frustrated, go back. The version you tried and the version that exists today are almost unrecognizable from each other, and the version that exists tomorrow will probably be better than both.

Terminal Access: For When You Need Claude Code

Most of you will live in Cowork and never need to think about this section. But if you’re on a version of Windows that won’t run the desktop app, or if you want to graduate to Vibe coding with Claude Code (the more powerful sibling that can connect to your servers, databases, and basically anything with a login), you’ll need to use what’s called a terminal.

Quick aside: Yes, there is a version of Claude Code on the Desktop App. Yes, it’s pretty good, and you can use it for a lot of cool things. But not #allthethings. So if you’re not yet able to use Cowork, or want to unlock Code’s full potential, read on for hooking up the version that’s called Claude CLI (Command Line Interface).

For the GenX kids: you already know what this is, even if the word “terminal” doesn’t ring a bell. Remember DOS? Remember that black screen with the blinking cursor where you typed C:\> dir to see your files and C:\> OREGON.EXE to die of dysentery on the trail to Oregon? That’s a terminal. It’s a text-based way to talk to your computer by typing commands instead of clicking on icons. Your Mac has one built in (it’s called Terminal, creatively enough), and Windows has a couple of options called Command Prompt and PowerShell.

I use an app called Warp instead of the built-in terminal, and I recommend it if you’re going down this road. Warp is a modern terminal that doesn’t look like it was designed in 1987 — it has autocomplete, it’s easier to read, and it doesn’t make you feel like you accidentally hacked into NORAD. It’s free, it works on Mac (and Windows support is available too), and it makes the Claude Code experience significantly less intimidating. It has a built-in AI you can chat with to ask it to help you do things like, “Hey, install Claude Code for me” and then “Be sure I can run Claude Code everywhere on my computer.” and “Hey, can you make me a latte?”

Give Anthropic like maybe 20 minutes and that one might be a reality.

If you get stuck, drop a message in the Discord and we’ll help you get sorted. Or grab one of the upcoming sessions where we’ll walk through it live.

Projects: Your First Move in the Desktop App

Once you have the desktop app installed, the first thing to do is create a Project. Projects are how you bundle related conversations, files, and context together so Claude remembers what you’re working on across multiple chats.

Think of it this way: without a project, every conversation with Claude starts from scratch. It doesn’t remember your last chat, your preferences, or what you told it about your book series yesterday. With a project, you create a container where Claude has ongoing context about a specific area of your work.

You might create a project for your book launch, another for your newsletter, another for your backlist marketing. Inside each project, you can attach reference documents — your brand guide, your blurb, your series bible, whatever Claude needs to give you useful answers without you re-explaining everything each time.

During the summit, I walked through how my own projects are organized, and the main takeaway was that projects are the difference between Claude being a generic chatbot and Claude being an executive assistant that actually knows your business. If you’re just getting started, skip regular chat entirely and go straight to creating one project for the area of your business where you spend the most time.

The Two Things That Make Cowork Powerful

Cowork on its own is already useful — you point it at a folder, describe what you want, and it gets to work. But the features that turn it from a nice-to-have into something that can genuinely run parts of your business are Skills and Connectors.

Connectors are how Claude talks to your other tools. When you go into Customize in the desktop app, you’ll see a directory of approved connectors — Google Drive, Gmail, Google Calendar, Notion, Slack, Stripe, Zapier, n8n, and a growing list that gets longer every week. Connecting one is as simple as clicking “add” and signing into your account. Once connected, Claude can read from and write to that service without you having to copy and paste between windows.

If you’ve been reading this newsletter for a while, you might remember my deep dives on MCP servers. Connectors are the friendly, click-to-install version of that same concept. MCP stands for Model Context Protocol (you don’t need to remember that), and it’s the bridge that lets Claude talk to your apps without needing to understand the technical details of each one. The official connectors in the directory have been vetted by Anthropic, so they’re the safest place to start. For those of you who want to go further and connect things that aren’t in the official directory (like authorautomations.social), you can set up personal connectors, which we covered extensively during the summit and which I’ll walk through in detail next week.

One limitation worth knowing: the official connectors currently only support one account per service. So if you have multiple Google accounts (I have approximately 4,000, which is a me problem), you’re limited to connecting one through the official connector. There are workarounds for this using personal connectors and Claude Code, and I’ll cover those in a future installment.

Skills are how you teach Claude to do things your way. A skill is basically a set of instructions — a recipe card, if you will — that tells Claude how to handle a specific type of task with your preferences, your voice, your tools, and your standards baked in.

I have over 100 skills at this point, which is admittedly excessive. I have a developmental editor skill, a line editor skill, a continuity specialist skill, a pen name architect skill, a series architect skill, a copywriting skill, and on and on. Each one contains specific instructions about how I want that task done, which tools to use, where to put the output, and what standards to apply.

You don’t need 100 skills. You need one. Pick the task you do most often — maybe it’s drafting social media posts, or summarizing your reader feedback, or formatting your newsletter — and start a conversation with Claude in Cowork. Say “interview me about how I do this task, what tools I use, and how I want the output to look.” Claude will ask you questions, and when you’re done, you’ll have a skill you can save and reuse every time you do that task.

The beautiful part about skills (and this came up several times during the summit) is that you don’t have to go back into the settings menu to update them. If you’re chatting with Cowork throughout the day and you realize the skill is missing something, just say “hey, add this to my copywriting skill as a non-negotiable rule.” Claude updates the skill for you. You’re teaching it incrementally, the same way you’d train a real assistant — one correction at a time, in the flow of actual work.

Where This Series Is Headed

Next week, I’m going deeper on Skills and Connectors with specific walkthroughs for authors. I’ll show you how to set up your first connector (we’ll start with Google Drive since nearly everyone uses it), how to create a skill from scratch by having Claude interview you, and how the two work together to turn Cowork from a fancy chatbot into something that actually handles repeatable tasks in your business. I’ll also touch on a big one —how to manage some security and risk with connectors and skills.

If you came to the summit, next week’s newsletter will feel like a practical follow-up to what you saw in the MCP Servers and Skills session. If you didn’t make it to the summit, you’ll have everything you need to get started.

Reply and tell me: what’s the first task you’d want to teach Claude to do your way? I read every one.

—Chelle

Claude for Authors, Part 2: Getting Set Up…